Analysis of the Molecular Weight Distribution of Petroleum

for Energy Utilization, National Institute of Advanced Industrial Science and Technology, 16-1 Onogawa, Tsukuba 305-8569, Japan, and Chemistry Div...
0 downloads 0 Views 215KB Size
Energy & Fuels 2004, 18, 1405-1413

1405

Analysis of the Molecular Weight Distribution of Petroleum Asphaltenes Using Laser Desorption-Mass Spectrometry Ryuzo Tanaka,*,† Shinya Sato,‡ Toshimasa Takanohashi,‡ Jerry E. Hunt,§ and Randall E. Winans§ Central Research Laboratories, Idemitsu Kosan Co., Ltd., 1280 Kamiizumi, Sodegaura 299-0293, Japan, Institute for Energy Utilization, National Institute of Advanced Industrial Science and Technology, 16-1 Onogawa, Tsukuba 305-8569, Japan, and Chemistry Division, Argonne National Laboratory, 9700 South Cass Avenue, Argonne, Illinois 60439 Received November 5, 2003. Revised Manuscript Received June 1, 2004

Laser desorption mass spectrometry (LD-MS) was used to measure three asphaltenes, their sub-fractions separated using gel-permeation chromatography (GPC), and model compounds to ascertain their molecular weight distributions and averaged molecular weights. The asphaltenes were isolated from three crude oils: Maya (MY), Khafji (KF), and Iranian Light (IL). To optimize the measurement conditions for LD-MS, the effects of instrument mode, matrix, and laser energy were determined. Linear mode detection gave larger ion counts and higher signal-to-noise ratios than did reflector mode. The use of a matrix to determine the asphaltene molecular weight distribution was not useful as it enhanced the ionization of only some of the asphaltene fractions. The chosen laser energy provided significant ionization of the high-molecular-weight fractions, while minimizing polymerization and fragmentation. Various model compounds were measured to clarify the dependence of the ionization capacity on molecular structure. Pure and small alkylsubstituted aromatics ionize easily, and tend to polymerize at relatively low laser energies. Bridged aromatics are easily disrupted at the alkyl bridge and do not polymerize. Aliphatic hydrocarbons are very difficult to ionize in the absence of an appropriate matrix. The higher the molecular weight of a compound, the more difficult it is to ionize. The characterization of the asphaltenes was aided by separating each into six sub-fractions using GPC. These fractions fell into two groups: high- and low-aromaticity groups. In each group, the lower the molecular weight of a sub-fraction, the higher its aromaticity. Maya asphaltene (As-MY) had a larger molecular weight fraction than the other asphaltenes, but the distribution pattern of As-KF was similar to that of As-MY, except for this fraction. All the As-IL sub-fractions had lower molecular weights and higher aromaticity than the corresponding sub-fractions of the other asphaltenes. The averaged molecular weights of the asphaltenes were determined from the weighted average of the averaged molecular weight of each of the six sub-fractions. For As-MY, As-KF, and As-IL the values are 1657, 1628, and 1462 amu, respectively.

Introduction One effective way to reduce both energy consumption and CO2 emissions is to improve the efficiency of the processes used to upgrade heavy oil. In refinery operations, the processing of the heavy asphaltene-rich fraction can cause serious problems, such as rapid deactivation of catalysts, coking within the reactor or fractionator, and formation of sludge in the products. An understanding and appreciation of the chemistry and physics of asphaltenes is vital to control this refractory fraction and to improve the efficiency of these processes. Asphaltenes are the most polar and heaviest compounds in crude petroleum. They are a closely related group of compounds that associate in solution to form * Corresponding author. Tel: +81 (438) 75-4380. Fax: +81 (438) 75-7213. E-mail: [email protected]. † Idemitsu Kosan Co., Ltd. ‡ National Institute of Advanced Industrial Science and Technology. § Argonne National Laboratory.

complex colloidal structures. To moderate asphaltene behavior, it is essential to understand their molecular structure, associative properties, and reactivities, and the relationships among them. Research in this field has been reviewed in several papers.1-5 The molecular structures of asphaltenes are poorly understood. They are very complex mixtures. There is no consensus on the molecular weight distributions or average molecular weight of different fractions, both of which are fundamental for determining their molecular structure. Strausz et al. measured the average molecular weight of Athabasca asphaltene as 6000 amu using (1) Speight, J. G. The Chemistry and Technology of Petroleum; Marcel Dekker: New York, 1980. (2) Chemistry of Asphaltenes; Bunger, J. W., Li, N. C., Eds.; American Chemical Society: Washington, DC, 1984. (3) Asphaltenes, Fundamentals and Applications; Sheu, E. Y., Mullins, O. C., Eds.; Plenum Press: New York, 1995. (4) Structures and Dynamics of Asphaltenes; Mullins, O. C., Sheu, E. Y., Eds.; Plenum Press: New York, 1998. (5) Sheu, E. Y. Energy Fuels 2002, 16, 74-82.

10.1021/ef034083r CCC: $27.50 © 2004 American Chemical Society Published on Web 07/10/2004

1406 Energy & Fuels, Vol. 18, No. 5, 2004

the ruthenium-catalyzed oxidation reaction (RICO) and nuclear magnetic resonance (NMR).6 By contrast, Mullin et al. found it to be around 750 amu using UVfluorescence.7 Subsequently, a wide range of average molecular weights from 400 to 10000 amu have been reported using mass spectrometry, vapor pressure osmometry (VPO), and size-exclusion chromatography.8-10 Two reasons are given for the wide variety of asphaltene molecular weights. First, the molecular weights of asphaltenes vary widely between different source crude oils. Alternatively, the value is highly dependent on both the method and measurement conditions. Methods such as VPO or GPC, in which solvents allow self-association within the solvent, overestimate the asphaltene molecular weight. Conversely, molecular weight distributions measured using mass spectrometry may be inexact because of fragmentation or incomplete ionization. Laser desorption-mass spectrometry (LD-MS) is considered one of the better methods of obtaining the comprehensive molecular weight distribution of an asphaltene.11,12 It possesses the potential to ionize asphaltene molecules across a range of molecular weights although, like simple mass spectrometry, it can be affected by fragmentation, polymerization, and ionization efficiency. This study examined the optimum conditions for measuring asphaltene molecular weights using LD-MS, and ascertained the molecular weight distributions of three petroleum asphaltenes and their GPC subfractions. The asphaltenes were isolated from three different crude oils: Maya (a Mexican heavy crude oil), Khafji (an Arabian heavy), and Iranian light (an Iranian intermediate). LD-MS of model compounds was also measured in order to establish the relationship between molecular structure and the ionization capacity, fragmentation, and polymerization of asphaltenes. Experimental Sample Preparation. Asphaltene Extraction. The residua after vacuum distillation of three crude oils at >500 °C were obtained. The asphaltenes were isolated by adding a 20:1 excess of n-heptane to each residue at 25 °C. The suspension was stirred for 1 h at 100 °C in an autoclave, and then cooled, filtered, and allowed to stand at 25 °C overnight. The precipitate was washed twice with n-heptane and dried. The yields of asphaltenes (precipitates) of Maya, Khafji, and Iranian Light were 24.9, 14.2, and 6.3 wt % respectively. Fractionation with GPC. Each asphaltene was further separated using GPC to determine the ionization differences between the lighter and heavier fractions. The details of the separation procedures are given elsewhere.13 In brief, 0.5 mL of a 2 wt % chloroform solution of asphaltene were injected along with chloroform at a rate of 3.0 mL/min. The eluent was initially collected as 30 sub-fractions that were subsequently pooled to give six sub-fractions for each sample. (6) Strausz, O. P.; Mojelsky, T. W.; Lown, E. M. Fuel 1992, 71, 13551363. (7) Groenzin, H.; Mullins, O. C. Energy Fuels 2000, 14, 677-684. (8) Miller, J. T.; Fisher, R. B.; Thiyagarajan, P.; Winans, R. E.; Hunt, J. E. Energy Fuels 1998, 12, 1290-1298. (9) Boduszynski, M. W. Chemistry of Asphaltenes; Bunger, J. W., Li, N. C., Eds.; American Chemical Society: Washington, DC, 1984; Chapter 7. (10) Anderson, S. I. Fuel Sci. Technol. Int. 1994, 12, 51. (11) Yang, M.-G.; Eser, S. ACS Reprints, ACS New Orleans Meeting, 768, 1999. (12) Fujii, M.; Yoneda, T.; Sato, M.; Sanada, Y. J. Jpn. Pet. Inst. 2000, 43, 149-156. (13) Sato, S.; Takanohasi, T.; Tanaka, R. Prepr. Pap.sAm. Chem. Soc., Div. Fuel Chem. 2003, 48 (1), 61-62.

Tanaka et al. Table 1. Properties of Intact Asphaltenes and Their Sub-fractions’ Yields sample

MY

KF

IL

yields in VR, wt % elemental analysis, wt % C H N S O metals, wt ppm Ni V atomic ratio H/C N/C S/C O/C Mn (VPO) fa (13C NMR)

24.9

14.2

6.3

82.0 7.5 7.1 1.3 1.2

82.2 7.6 7.6 0.9 1.1

83.2 6.8 5.9 1.4 1.5

390 1800

200 550

390 1200

1.09 0.07 0.01 0.01 4000 0.53

1.10 0.08 0.00 0.01 4000 0.51

0.97 0.06 0.01 0.01 2400 0.54

sub-fraction yields, wt % Fr-1 Fr-2 Fr-3 Fr-4 Fr-5 Fr-6

17.5 12.7 13.1 13.4 12.3 31.0

16.7 10.8 10.4 10.3 12.9 38.8

16.9 17.2 17.3 16.2 14.5 18.0

Analysis. Ultimate Analysis and VPO. The intact asphaltenes and sub-fractions were subjected to the same elemental analysis. The carbon and hydrogen contents of each asphaltene were measured using a CHN600 (LECO); the sulfur, nitrogen, and oxygen contents were determined using an AQS-6W (Tanaka Scientific Instrument), an ANTEK7000 (Antek), and a varioEL III (Elementar), respectively. The metal contents were analyzed using inductively coupled plasma spectrophotometry (ICP) using an SPS1500VR spectrometer (Seiko Instruments). The densities were determined in conformity with JIS K 7112 using a DMA45 (Paar), and the molecular weights were measured using an Automatic Molecular Weight Apparatus (Rigosha). Mass Spectrometry. LD-MS measurements were carried out with a Voyager DE TSR (Applied Biosystems). The asphaltenes, their sub-fractions, and the model compounds, with and without matrix, were dissolved completely in chloroform using a vibrator to give 5 wt % solutions. A 1-mL aliquot of each solution was placed in a gold-lined sample holder and the chloroform evaporated in the air. The sample holder was then placed in the spectrometer and measurements were made under various conditions.

Results and Discussion Asphaltene Properties. The properties of each asphaltene are given in Table 1, along with the yields of the different sub-fractions. Maya asphaltene (As-MY) is the heaviest. It has the greatest density and contains the most metals of the three asphaltenes. Khafji asphaltene (As-KF) is moderately heavy and contains the fewest metals. The H/C atomic ratio of As-KF is the highest, consistent with this asphaltene having the lowest aromaticity. Iranian Light asphaltene (As-IL) is the lightest. It has the lowest density and smallest molecular weight, but contains considerable nitrogen and metals. It also has the lowest H/C atomic ratio, indicating that it possesses the highest aromaticity. As seen in Table 1, As-MY and As-KF did not differ markedly except for metals. For example, their hydrogen contents were similar at 7.5 and 7.6. Nonetheless, the degree of aggregation of these two asphaltenes

MW Distribution of Petroleum Asphaltenes

Figure 1. LD-MS spectra of (a) maltene and (b) asphaltene from Maya.

differs markedly, on the basis of small-angle neutron scattering14 and their coking reactivity determined in autoclave experiments.15 As-MY tends to aggregate more and to coke more than As-KF. In a recent thermal analytical study, Zhang et al. found that metals influence coke yield minimally.16 Hence, the two asphaltenes must possess distinct differences in their molecular structure that result in their very different properties. The aromaticity (fa) of the two asphaltenes is 0.53 and 0.51. The 0.2 difference in fa is not negligible, as coke yields are sensitive to fa. Gray et al. demonstrated that in thermal cracking of asphaltenes coke yields are dominated by the carbon structure of the asphaltenes and linearly proportional to the fa.17 The yields of Fractions 1-5 are similar, while that of Fraction 6 is higher in all three asphaltenes. LD-MS Spectra of Maltene and Asphaltene. Figure 1 shows the LD-MS spectra of maltene and the asphaltene extracted from Maya vacuum residue. The molecular weight distribution of the maltene peaks at about 400 amu. By contrast, the asphaltene has a bimodal distribution with peaks near 500 and 2000 (14) Tanaka, R.; Hunt, J. E.; Winans, R. E.; Thiyagarajan, P.; Sato, S.; Takanohashi, T. Energy Fuels 2003, 17, 127-134. (15) Tanaka, R.; Sato, S.; Takanohashi, T.; Sugita, S. Prepr. Pap.s Am. Chem. Soc., Div. Fuel Chem. 2003, 48 (1), 98-99. (16) Zhang, Y. Personal communication, 2003. (17) Rahmani, S.; McCaffrey, W. C.; Dettman, H. D.; Gray, M. R. Energy Fuels 2003, 17, 1048-1056.

Energy & Fuels, Vol. 18, No. 5, 2004 1407

Figure 2. LD-MS spectra of As-MY, (a) in reflector mode without matrix, (b) in reflector mode with a dithranol matrix, (c) in linear mode without matrix.

amu. The molecular weight distribution of maltene is relatively stable irrespective of measurement conditions, while that of asphaltene is heavily dependent on conditions. A series of experiments were conducted to ascertain the influence of instrument mode (linear or reflector), matrix, and laser energy on the asphaltene molecular weight distribution. Optimizing Measurement Conditions. Instrument Mode and Matrix. Linear mode uses a linear detector only. It is the most sensitive mode, owing to the shorter flight path. It is also the more sensitive, given that fragments, neutrals, and molecular ions arrive at the detector at the same time. Reflector mode uses a reflector detector only and has a higher resolution and greater mass accuracy owing to the longer flight path and the action of the reflector.18 For weakly ionized samples, the ionization efficiency may be drastically improved with an appropriate matrixsa technique called matrix-assisted laser desorption ionization (MALDI). Although the mechanism is not fully understood, it is thought that the initial ionization of the matrix assists in subsequent ionization of the sample. In this study, the matrix effect of dithranol, a substance commonly used for high-molecular-weight samples, was tested. Figure 2 shows the spectra of As-MY in (a) reflector mode without matrix, (b) reflector with matrix, (18) Voyager Biospectrometry Workstation User’s Guide, PerSeptive Biosystems, Inc., Framingham, MA, 1996.

1408 Energy & Fuels, Vol. 18, No. 5, 2004

Tanaka et al.

Figure 4. The relationship between laser steps and the energy of the laser pulse.

Figure 3. LD-MS spectra of As-MY collected with laser energies of (a) 1900, (b) 2100, and (c) 2300 LS.

and (c) linear without matrix. The reflector mode spectrum (a) has a lower signal-to-noise ratio and lesswell-defined peaks in the higher-molecular-weight range than does the linear mode spectrum (c), demonstrating that reflector mode is insufficiently sensitive at higher molecular weights. Figure 2(b) shows that the addition of a matrix enhances the lower-molecular-weight component peaks considerably, either via the action of dithranol itself, or its derivatives, or some smaller molecules in the asphaltene. Asphaltene is a complex mixture of a large number of components with a variety of molecular structures. Some constituents appear to be ionized selectively in the presence of dithranol. However, it is inappropriate to undertake molecular weight distribution analyses of asphaltenes by selective ionization using a matrix. Consequently, all subsequent measurements were made in linear mode without matrix, using simple laser desorption ionization (LDI). Laser Energy. The criteria used to optimize the laser energy during ionization require that all components up to and including the heaviest become significantly ionized, while fragmentation and polymerization are insignificant. Figure 3 shows the laser spectra of AsMY collected at three energy levels: (a) 1900, (b) 2100, and (c) 2300 laser steps (LS). The relationship between the laser power increment with the unit used in the experiments and the energy of the laser pulse striking the sample, expressed in µJ, is shown in Figure 4. It is

not practicable to derive an absolute value for this relationship because of differences between the laser models. The energy of the laser pulse striking the sample is very sensitive to differences in the optical system and aging of the laser, with a consequent decline in the emitted energy. For As-MY, the higher-molecularweight fractions do not appear to be ionized significantly at a laser energy of 1900 LS. By contrast, at 2100 LS, more high-molecular-weight fractions are present and fragmented ions with low molecular weights, around 200 amu, are not significant. Therefore, measurements of asphaltene should be made with a laser energy of more than 2100 LS. Ionization Efficiency of the Model Compounds. The LD-MS values of some model compounds were measured to explore the relationship between molecular structure and ionization capacity. Structural features of interest included aromatic rings, alkyl side chains and bridges, and heteroatoms. Figure 5 shows the LD-MS spectra of dibenzopyrene (MW ) 302) collected using laser energies of (a) 1700, (b) 2000, and (c) 2400 LS. Dibenzopyrene readily ionizes at around 1700 LS, and polymerization still occurs at 2000 LS. The peaks at 604 and 906 amu in Figure 5(b) were assigned to the dimer and trimer ions. At 2400 LS, tetramer peaks were detected at about 1208 amu, and fragmented ions were observed in regions below 302 amu. In Figure 6, the ion counts of prime, polymer, and fragmented ions of dibenzopyrene integrated in the mass regions are 300-314, 314-1600, and 0-300 amu, respectively. In Figure 6, the ordinate plots the integrated ion count on a logarithmic scale and the abscissa plots the laser energy in LS. Prime and polymer ions dominate at 1700 LS, while the proportions of polymer and fragmented ions increase with the laser energy. At 2400 LS, fragmented ions exceed prime ions. Figure 7 shows the LD-MS spectra of 2-ethylanthracene at 1200-2200 LS. No signals were detected at 1200 LS. Only the prime ion peak at 206 amu occurs at 1400 LS. The same peak is present at 1600 LS, along with that of the dimer ion at 412 amu. Peaks from fragments at 40 amu and other mass numbers less than 202 amu begin to appear at 1800 LS. Figure 8 shows the integrated ion counts of prime, polymer, and fragment ions of 2-ethylanthracene for the integrated mass

MW Distribution of Petroleum Asphaltenes

Figure 5. LD-MS spectra of dibenzopyrene at laser energies of (a) 1700, (b) 2000, and (c) 2400 LS.

Figure 6. Integrated ion counts of prime, polymer, and fragmented ions in LD-MS measurement of dibenzopyrene.

regions 202-206, 206-800, and 0-202 amu, respectively. The intensities of all the ion peaks at 1200 LS and of the polymer and fragmented ion peaks at 1400 LS are below the noise level of less than 20 counts. Consequently, integrated ion counts of each group of ions with counts of 50 are taken as zero on the ordinate. The figure shows the quantitative trend with the prime ion peaks dominating at 1400 and 1600 amu. The polymer and fragmented ion peaks become enhanced with increasing laser energy. These measurements of dibenzopyrene and 2-ethylanthracene showed that pure and little alkyl-substi-

Energy & Fuels, Vol. 18, No. 5, 2004 1409

Figure 7. LD-MS spectra of 2-ethylanthracene at laser energies of (a) 1200, (b) 1400, (c) 1600, (d) 1800, (e) 2000, and (e) 2200 LS.

Figure 8. Integrated ion counts of prime, polymer, and fragmented ions in LD-MS measurement of 2-ethylanthracene.

tuted condensed-aromatic hydrocarbons are relatively easy to ionize as compared to asphaltenes, which need more than 2100 LS to produce any degree of ionization, but which also begin to polymerize at relatively lower laser energies. Figure 9 shows the LD-MS spectra of rubrene at 2000 and 2400 LS. At 2000 LS, the prime ion peak of 532 amu dominates and small polymer ion peaks near 1100 amu appear. At 2400 LS, large peaks at 376 and 454 amu, arising from fragmented ion peaks, dominate along with other smaller peaks. A distinctive dimer peak appears at 1064 amu. The two fragmented ion peaks have masses of 78 and 156 amu, consistent with the

1410 Energy & Fuels, Vol. 18, No. 5, 2004

Figure 9. LD-MS spectra of rubrene at laser energies of (a) 2000 and (b) 2400 LS.

Figure 10. Integrated ion counts of prime, polymer, and fragmented ions in LD-MS measurement of rubrene.

removal of one or two benzene rings from rubrene. The trends for each ion peak are quantified in Figure 10 as integrated ion counts over the mass regions 526-540, 540-1600, and 0-526 amu for prime, polymer, and fragment ions, respectively. In contrast to the pure or little alkyl-substituted condensed aromatic hydrocarbons in which the polymer ion peaks dominate over the fragmented ion peaks, bridged-type aromatic hydrocarbons show many fragmented ion peaks, since their bridge structures are easier to disrupt with laser desorption ionization. Peaks arising from polymer ions tend to be less prominent.

Tanaka et al.

Figure 11. LD-MS spectra of PEGs in a matrix of 2,5dihydroxybenzoic acid: (a) PEG600 and (b) PEG1500.

As representative non-aromatic compounds, two poly(ethylene glycol)s (PEGs) were examined: PEG600 and PEG1500, mixtures of PEG with molecular weights around 600 and 1500 amu, respectively. Non-aromatics such as PEG are difficult to ionize by laser desorption in the absence of a matrix, but in the presence of a matrix, such as 2,5-dihydroxybenzoic acid, appropriate molecular weight distributions appear. Figure 11 shows the LD-MS spectra of the two PEGs in such a matrix. Along with the main PEG peaks around 600 and 1500 amu, matrix and fragment ion peaks occur at about 200 amu. Figure 12 shows the LD-MS spectrum of the solar dyne, N,N′-bis(2,5-di-tert-butylphenyl)-3,4,9,10-perylenedicarboximide. The structure of this dyne is considered most akin to that of asphaltenes of the model compounds tested here. It has a relatively higher molecular weight of 766 amu and a complex structure that includes polyaromatic rings, alkyl side chains, and heteroatoms. The dyne was not ionized at a laser energy of 2400 LS or below, but started to ionize at 2700 LS, with some fragmentation, but little polymerization. In summary, these results indicate that pure and small alkyl-substituted poly-nuclear aromatics are easily ionized. They start to polymerize at relatively low laser energies and continue to polymerize at higher laser energies. Bridged aromatics are ionized at relatively low laser energy, but start to fragment at moderate laser energy. Non-aromatics are difficult to ionize using laser desorption, except in the presence of an appropriate matrix. A dyne with a structure akin to asphaltenes in

MW Distribution of Petroleum Asphaltenes

Figure 12. LD-MS spectra of N,N′-bis(2,5-di-tert-butylphenyl)-3,4,9,10-perylenedicarboximide at laser energies of (a) 2400 and (b) 2700 LS.

having aromatic rings, alkyl side chains, and heteroelements, is ionized only with relatively high laser energy, resulting in fragmentation and limited polymerization. Measurement of GPC Sub-fractions of Asphaltenes. Molecular Weight Distribution of Sub-fractions. Figure 13 shows the LD-MS measurements of each of the As-KF sub-fractions. The spectra were obtained using automatic measurements in which the optimum combination of sample position and laser step for each sample were found in a series of trial measurements made before data acquisition.14 The sub-fractions from all three asphaltenes share spectral features. Fractions 1, 2, and 3 have broad molecular weight distributions centered near 1800 amu and ranging from 800 to 4000 amu. Fractions 4 and 5 have relatively narrow distributions centered on 400 to 800 amu. Fraction 6 has a broad distribution pattern closely resembling those of the intact asphaltenes. Ionization Capacity and Molecular Structure of Subfractions. The properties of the GPC sub-fractions were investigated using ultimate analysis and 13C NMR. The number-averaged molecular weight (Mn) of the subfractions was estimated using GPC and LD-MS, using LD-MS measurements to ascertain the laser energy dependence of Mn. The results are shown in Table 2. The sub-fractions of each asphaltene share features and can be classified into two groups using aromaticity. Fractions 1-4 have relative low aromaticity (fa < 0.52), whereas Fractions 5 and 6 have relatively high aroma-

Energy & Fuels, Vol. 18, No. 5, 2004 1411

Figure 13. LD-MS spectra of As-KF sub-fractions: Fractions (a) 1, (b) 2, (c) 3, (d) 4, (e) 5, and (f) 6.

ticity (fa > 0.52). In each group, the higher the aromaticity of a sub-fraction, the lower is its molecular weight. The average molecular weights of sub-fractions have maxima at 2300 LS, perhaps because the high-molecular-weight portions fail to ionize sufficiently at lower laser energies, while fragmentation becomes more pronounced at higher energies. Consequently, in this study, 2300 LS was chosen as the best laser energy for laser desorption ionization of all the asphaltene sub-fractions. Note that the optimized laser power is highly dependent on the optical system of each apparatus and on the age of the laser unit, as well as on sample preparation, including both the number of fractions and separation methods, and on the number of measurements. Each individual asphaltene constituent may have its own optimized conditions; the greater the number of subfractions and measurements used, the more precisely the conditions for measuring each sub-fraction can be determined. The distinct features of each asphaltene are shown in Figure 14, which plots the values of Mn at 2300 LS with the fa of each sub-fraction, and the areas of the plotted circles are proportional to the yield of each subfraction. As-MY has a higher-molecular-weight portion (Fraction 1) than other asphaltenes, but otherwise the overall distribution of As-KF is similar to that of AsMY. All of the As-IL fractions plot in the lowermolecular-weight and higher aromaticity area, especially Fraction 5, which has a very high aromaticity with fa ) 6.2.

1412 Energy & Fuels, Vol. 18, No. 5, 2004

Tanaka et al.

Table 2. Properties of Asphaltene Sub-fractions asphaltene fraction

MY 1

2

ultimate analysis, wt % C 79.37 81.52 H 7.88 7.91 N 1.18 1.20 S 7.17 7.52 O(diff) 4.40 2.47 atomic ratio H/C 1.18 1.16 N/C 0.01 0.01 S/C 0.03 0.03 O/C 0.04 0.02 13 fa ( C NMR) 0.424 0.465 Mn (GPC) 6984 3318 Mn (LDMS) laser step 2000 1849 1715 laser step 2300 2169 1701 laser step 2600 1801 1443 laser step 2900 1697 -

KF 4

5

6

1

2

3

4

5

6

1

2

3

4

5

6

81.57 7.67 1.39 6.90 2.37

83.19 7.47 1.63 5.34 3.10

81.63 6.98 1.60 6.70 5.57

80.49 7.11 1.52 5.31 3.96

79.91 7.51 0.98 7.63 2.47

81.26 7.75 0.95 7.57 2.64

81.34 7.69 0.96 7.37 1.99

82.19 7.43 1.03 7.36 3.89

80.43 6.80 1.08 7.80 9.35

75.50 7.22 1.08 6.85 4.74

80.80 7.75 1.46 5.25 2.57

82.78 7.75 1.48 5.42 2.52

82.89 7.81 1.45 5.33 2.37

83.19 7.47 1.63 5.34 3.33

82.74 6.80 1.74 5.38 4.87

79.52 7.07 1.42 7.13 4.74

1.12 0.02 0.03 0.02 0.448 1677

1.07 0.02 0.02 0.02 0.461 674

1.02 0.02 0.03 0.03 0.549 145

1.05 0.01 0.02 0.05 0.520 3165

1.12 0.01 0.04 0.04 0.437 7856

1.14 0.01 0.03 0.02 0.438 3617

1.13 0.01 0.03 0.02 0.438 1712

1.08 0.01 0.03 0.02 0.473 725

1.01 0.01 0.04 0.04 0.547 221

1.14 0.02 0.03 0.09 0.530 2230

1.14 0.02 0.02 0.04 0.467 5682

1.12 0.01 0.02 0.02 0.457 3029

1.12 0.02 0.02 0.02 0.478 1608

1.07 0.02 0.02 0.02 0.521 658

0.98 0.02 0.02 0.03 0.606 189

1.06 0.02 0.03 0.05 0.550 4735

1971 1595 1437 -

634 1214 1279 -

497 1366 1310 -

1645 1682 1527 -

1823 1530 -

1741 1429 -

1578 1398 -

710 1288 1346 -

512 1336 1300 -

1714 1556 -

1699 1647 -

1571 1409 -

1400 1307 -

711 1123 1182 -

475 1272 1190 -

1653 1493 -

Figure 14. Averaged molecular weight and aromaticity of asphaltene sub-fractions. Table 3. Averaged Molecular Weight of Asphaltenes Determined by GPC, VPO, and LD-MS MY sample

whole

VPO GPC LD-MS

4000 873 1147

IL

3

KF

average

whole

2953 1657

4000 1049 1347

Conversely, the order of the averaged molecular weight from the sub-fractions determined using GPC and LD-MS was consistent, with MY g KF > IL. In this instance, any differences arising from interactions in the GPC owing to the lower-molecular-weight and high aromaticity fractions, as compared to the other asphaltene fractions, are minimized because the fractions are measured separately. In the LD-MS measurements, differences in the fraction ionization capacities are minimized by applying weighted averages between the yields. Of the three methods, LD-MS is thought to provide a superior picture of the molecular weight distribution in asphaltenes as it avoids the effects of asphaltene molecular aggregations that occur in solvents. Consequently, the yield-weighted averaged value of the averaged molecular weight of the sub-fractions measured using LD-MS can be taken as the averaged molecular weight of the asphaltene. For As-MY, As-KF, and AsIL, this is 1657, 1628, and 1462 amu, respectively. Conclusion

IL

average

whole

average

2850 1628

2400 1608 1326

2741 1462

Averaged Molecular Weights of the Intact Asphaltenes. Table 3 shows the averaged molecular weights obtained using VPO, GPC, and LD-MS for each of the intact asphaltenes, i.e., before fractionation with GPC. The averages are the yield-weighted averaged values of the averaged molecular weight of each sub-fraction. The order of the averaged molecular weights from the intact asphaltenes determined using VPO, GPC, and LD-MS was MY ) KF > IL, IL > KF > MY, and KF > IL > MY, respectively. In particular, the averaged molecular weight of As-MY determined using GPC is extremely low. These results reflect differences in the instrumental and analytical procedures. In GPC measurements, the lower-molecular-weight and high aromaticity fractions of the asphaltene interact to a higher degree with the gel in the column than do the other fractions, which increases the retention time. In LD-MS measurements, the lower-molecular-weight and high aromaticity fractions are ionized most readily and are relatively enhanced in a spectrum.

An LD-MS study of three different asphaltenes and their sub-fractions, along with analysis of relevant model compounds, ascertained common structural features of the asphaltenes. Each showed a continuous bimodal molecular weight distribution with maxima around 500 and 1800 amu, which reflects the presence of two main groups of constituents, one with a relatively high aromaticity (fa g 0.52) and the other a relatively low one (fa e 0.52). Each group possesses the same qualitative relationship between molecular weight and aromaticity: the higher the molecular weight of a given asphaltene sub-fraction, the lower its aromaticity. The averaged molecular weight of the three intact asphaltenes, i.e., before fractionation, was around 1500 amu; the individual values of As-MY, As-KF, and As-IL were 1657, 1628, and 1462 amu, respectively. A two-step procedure is recommended for determining the averaged molecular weight of an asphaltene. First, the number-averaged molecular weight of the six asphaltene sub-fractions is found using LD-MS. Second, a weighted average is calculated using the sub-fraction yields. This procedure minimizes the effects of differences in the ionization capacities of the asphaltene constituents.

MW Distribution of Petroleum Asphaltenes

Acknowledgment. This study was supported by a grant from the International Joint Research Grant Program from the New Energy and Industrial Technology Development Organization (NEDO). It was carried out in collaboration with the Central Research Laboratory, Idemitsu Kosan Co., Ltd., the National Institute of Advanced Industrial Science and Technology (AIST), and the Argonne National Laboratory, The Pennsylvania State University, the University of Alberta, and

Energy & Fuels, Vol. 18, No. 5, 2004 1413

Hokkaido University. The authors are grateful to Professor Isao Mochida of Kyushu University for his invaluable advice covering many aspects of asphaltene chemistry. The authors also acknowledge Mr. Hidenori Torii, Central Research Laboratories of Idemitsu Kosan Co., Ltd., for his informative suggestions and helpful support in LD-MS measurements. EF034083R